CN104081415A - Method, apparatus and computer readable recording medium for managing a reference face database to improve face recognition performance under a restricted memory environment - Google Patents
Method, apparatus and computer readable recording medium for managing a reference face database to improve face recognition performance under a restricted memory environment Download PDFInfo
- Publication number
- CN104081415A CN104081415A CN201280058022.2A CN201280058022A CN104081415A CN 104081415 A CN104081415 A CN 104081415A CN 201280058022 A CN201280058022 A CN 201280058022A CN 104081415 A CN104081415 A CN 104081415A
- Authority
- CN
- China
- Prior art keywords
- face image
- face
- reference surface
- registration
- data storehouse
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/10—Human or animal bodies, e.g. vehicle occupants or pedestrians; Body parts, e.g. hands
- G06V40/16—Human faces, e.g. facial parts, sketches or expressions
- G06V40/172—Classification, e.g. identification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/51—Indexing; Data structures therefor; Storage structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/50—Maintenance of biometric data or enrolment thereof
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Human Computer Interaction (AREA)
- Oral & Maxillofacial Surgery (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Collating Specific Patterns (AREA)
- Processing Or Creating Images (AREA)
- Image Processing (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
An embodiment of the present invention presents a method for managing a reference face database to improve face recognition performance under a restricted memory environment, said method comprising: a step of acquiring a new input face image; a step of determining a classification condition corresponding to the input face image with reference to classification conditions of existing registered face images of a reference face database used for face recognition; and a step of, if the classification condition corresponding to the input face image is a specific classification condition and the value of the face image corresponding to the specific classification condition in the reference face database is less than a preset threshold value, selecting the input face image as an image which can be additionally registered to the reference face database.
Description
Technical field
The disclosure relates to a kind of method, device and computer readable recording medium storing program for performing use facial recognition from digital picture pick device (digital camera for additionally registering and delete face image to improve, digital camera, camera phones etc.) discrimination of the face that comprises in the content of multimedia (picture or film image) that obtains, and more specifically, additionally register and delete face image with select will be in database additionally registration face image and replace previously stored face image to improve the performance of the facial recognition while using the digital picture pick device pictures taken with limited storage resources with this face image.
Background technology
Synchronize with the expansion in the camera phones market that comprises smart phone, digital camera, digital camera etc., had many development for the technology of picture shooting or mobile image.One of these technology are that the face of the people for identifying the image that uses the shooting of digital picture pick device is to automatically focus or the technology to this person's automatic camera in the time that this person makes specific Facial expression to this face.
But, in conventional environment, the hydraulic performance decline reason when face image of taking from diversity angle, multiple Facial expression and the illumination (natural light, artificial light etc.) from multiple direction is identification face.In the work of doing for the difficult problem overcoming above, in database, registration, for the basic reference surface hole pattern picture of facial recognition, can obtain the raising in performance by registering in advance diversity face image as much as possible.But, because as limited in the storage resources of the digital picture pick device of digital camera or camera phones, exist and do not have enough resources to store the problem of all available face images.
Correspondingly, need a kind of high efficiency method for recently taking or the face image of input is selected and registers the face image of performance operating effectively improving facial recognition, and in view of limited memory resource, delete more previously stored face images.
Summary of the invention
Technical matters
Therefore, object of the present disclosure is to address the above problem.
In addition, even if the face image of high facial recognition rate is still provided under the environment of the variation such as attitude, Facial expression, lighting angle when another object of the present disclosure is to provide a kind of method to utilize camera to take face image for selecting and being registered in.
Another object of the present disclosure is to provide a kind of method for deleting for the most unwanted registered face of identification face in the middle of the face image in previously registration, to use efficiently the limited memory resource of digital picture pick device.
Technical scheme
The representative configuration of the present disclosure for realizing above-mentioned purpose as follows.
According to a disclosed aspect, aspect, provide a kind of reference surface hole data storehouse of managing for improve the method for facial recognition performance in limited memory environments.The method comprises obtains new input face hole pattern picture; The class condition of the face image of previously having registered with reference to the reference surface hole data storehouse for facial recognition is determined the class condition with input face hole image correlation; And if to the class condition of input face hole image correlation be specific classification condition and with this specific classification condition relevant face image not as the predetermined threshold value in reference surface hole data storehouse abundant, select input face hole pattern picture as can be in reference surface hole data storehouse the image of registration additionally.
According to another aspect disclosed by the invention, provide a kind of for managing reference surface hole data storehouse for improve the device of facial recognition performance at limited memory environments.This device comprises for recording the reference surface hole data storehouse for the face image of facial recognition; And additional registration face selected cell, if this additional face selected cell of registering, for having obtained input face hole pattern picture, is determined with reference to the class condition of the face image of previously having registered in reference surface hole data storehouse and the class condition of new input face hole image correlation; And if to the class condition of input face hole image correlation be specific classification condition and with this specific classification condition relevant face image not as the predetermined threshold value in reference surface hole data storehouse abundant, select this input face hole pattern picture as can be in reference surface hole data storehouse the image of registration additionally.
In addition, be provided for realizing other method and apparatus of the present disclosure and for carrying out the computer readable recording medium storing program for performing of said method.
Technique effect
According to the disclosure, because can select the most effective face image and it is additionally registered for facial recognition in reference surface hole data storehouse, so can provide facial recognition with high discrimination in the time using digital picture pick device to take pictures to people.Moreover, can use expeditiously insufficient memory resource, because delete some least effectively images from the face image of previous registration, for will newly adding the face image in reference surface hole data storehouse to.
Brief description of the drawings
Fig. 1 is the schematic diagram illustrating according to the configuration of the digital device of disclosure embodiment.
Fig. 2 is the process flow diagram that the method for the face image that will additionally register according to the selection of embodiment of the present disclosure is shown.
Fig. 3 is the schematic diagram in reference surface hole data storehouse, and it illustrates the process of being selected the face image that will additionally register according to the method for embodiment of the present disclosure by digital device that uses.
Fig. 4 is the process flow diagram of the method for the face image that will additionally register according to the selection of another embodiment of the present disclosure.
Fig. 5 to 8 illustrates according to selecting in the face image of the previous registration from reference surface hole data storehouse of embodiment of the present disclosure and deleting for the not process flow diagram of the method for effective face image of identification face.
Embodiment
In below detailed description of the present disclosure, with reference to accompanying drawing, these accompanying drawings are illustrated and can be implemented specific embodiment of the present disclosure by the mode of explanation.These embodiment fully describe in detail, so that those skilled in the art can realize the disclosure.Should be appreciated that, although different, not necessarily mutually exclusive property of various embodiments of the present disclosure.For example, do not deviating under the prerequisite of the spirit and scope of the present invention, be combined in special characteristic described herein with an embodiment, structure and characteristics can be realized in further embodiments.In addition, should be appreciated that, do not deviating under the prerequisite of the spirit or scope of the present invention, the position of the Individual components in each disclosed embodiment or layout can be revised.Therefore, below describe in detail and should not be considered as on limited significance, and the scope of the present disclosure defines by appended claims, appended claims is together with suitably explaining with the full breadth of claims equivalence required for protection.In the accompanying drawings, similar reference number refers to the same or analogous function by diversified mode.
Those skilled in the art hereinafter, describe various embodiments of the present disclosure in detail with reference to accompanying drawing, so that can easily realize the disclosure.
In order to focus on the face of people or specific Facial expression in the time using camera to take pictures, first should be by the previously stored reference surface hole pattern picture comprising in the face of detection and database be relatively carried out to the people's who comprises in the captured image of detection and Identification face.But, if the reference surface hole pattern of facial recognition similarly be only take with Facial expression and same light according under the image of attitude (pose), facial recognition rate can be considerably low.This be because, although be stored in the image that comprises someone A in the reference surface hole pattern picture in database, if taken but the face of this person A is Facial expression, attitude and the direction of illumination different with Facial expression, attitude and the direction of illumination of the reference surface hole pattern picture from the same person of storing in database, likely is that in the image of taking, this person A cannot be correctly identified.Corresponding this person's multiple face image can be stored in database to strengthen facial recognition rate.But, existing problems, because portable set, the memory storage space of digital camera or camera phones is too little so that cannot store all multiple face images.
The disclosure proposes a kind of method, the method (a) is selected the most effective face image, and in reference surface hole data storehouse, additionally register selected face image for facial recognition, to strengthen the facial recognition rate while using finite memory storage space, and (b) from the face image of previous registration, delete some least effectively images, for will newly adding the face image in reference surface hole data storehouse to.The disclosure also proposes to apply the device of method above and for recording and realize the computer readable recording medium storing program for performing of the method.
Fig. 1 is the schematic diagram illustrating according to the configuration of the digital device 100 of disclosure embodiment.
With reference to figure 1, digital device 100 comprises reference surface hole data storehouse 110, adds registration face selected cell 120, registers in advance face replacement unit 130, communication unit 140 and control module 150.
According to embodiment of the present disclosure, reference surface hole data storehouse 110, additional registration face selected cell 120, in advance register face replacement unit 130, communication unit 140 and control module 150 one of them can be included in a bit the program module in digital device 100 or can communicate by letter with digital device 100.But Fig. 1 illustrates the reference surface hole data storehouse 110 that is all included in digital device 100, additional registration face selected cell 120, registers face replacement unit 130, communication unit 140 and control module 150 in advance.One of them of these program modules can adopt the form of operating system, application program or other program module to be included in digital device 100 a bit, and can physically be stored in multiple known storage arrangement.In addition one of them that, can these program modules is stored in the remote storage apparatus that can communicate by letter with digital device 100 a bit.This type of program module comprehensively comprises for after this describing but is not limited to this execution particular task or carries out the routine, subroutine, program, object, assembly, data structure etc. of particular abstract data type.
First, the data of each angle of the face of particular person can be recorded in reference surface hole data storehouse 110, as the photo of taking to the right rotation 45 degree from the positive photo of taking of people, face, the photo that face is taken to left side rotation 45 degree, photo, the photo that face is taken to left side 90-degree rotation etc. (referring to the after this description of Fig. 3) that face is taken to the right 90-degree rotation.Herein, previously mentioned angle only provides as example, and in addition, can also use the multiple amendment of three-dimensional extended, such as face photo, face that 30 degree the take photo etc. that 60 degree take that is inclined upwardly of upwards kicking up.
In addition, the data of the light of the each angle on the face that is incident upon particular person can be recorded in reference surface hole data storehouse 110, the photo taken the photograph from the photo-beat of people's front projection as utilized, utilize the photo taken the photograph from the photo-beat of the right projection of face, utilize the photo taken the photograph from the photo-beat of the left side projection of face, utilize photo that the photo-beat projecting takes the photograph from face below, the (not shown) such as photo that photo-beat that utilization projects above face is taken the photograph.Herein, previously mentioned angle only provides as example, and in addition, the multiple amendment of three-dimensional extended, it is also possible for example utilizing the photo of spending projection light from face upper left side 30, projection light is spent in utilization photo, the photo that utilizes the middle projection light between face front and the right etc. from face lower right 60.In addition, people's photo can be classified and record according to the type of used light source (indoor and outdoors light source, natural light, surround lighting etc.).
Next,, if input specific people's new input face hole pattern picture, additional registration face selected cell 120 can be carried out the function that determines whether to store new input face hole pattern picture in reference surface hole data storehouse 110.In this example, additional registration face selected cell 120 can also, after checking the face image of previously stored this specific people in reference surface hole data as above storehouse 110, determine whether in reference surface hole data storehouse 110, to store new input face hole pattern picture.
Fig. 2 is the process flow diagram of the method for the face image that will additionally register according to the selection of embodiment of the present disclosure.
In order to select to advise and to be registered as the face image of new input face hole pattern picture, providing a kind of parts according to the additional registration face selected cell 120 of embodiment of the present disclosure similarly is the no new registration condition with facial recognition for identifying new input face hole pattern, and in the case of determining that the condition of new input face hole pattern picture selects the candidate face of this new input face hole pattern picture as registration in reference surface hole data storehouse 110 satisfied.
With reference to figure 2, according to the first outside new face image that receives from reference surface hole data storehouse 110 of the additional registration face selected cell 120 of embodiment of the present disclosure, the previously face image (S201) of registration of reference surface hole data storehouse 110 storages.The new face image receiving from outside can be the new image of taking or previously be stored in the image external memory storage space.In addition, the data of even having stored in reference surface hole data storehouse 110 also can perform step S201, in the situation that not yet determining whether these type of data meet new registration condition, can so do.
Whether the condition of next, determining new input face hole pattern picture is the condition (S202) previously having existed in the face image of registration in reference surface hole data storehouse 110.
Here, the condition of new input face hole pattern picture is that feature based on input face hole pattern picture is by the standard of new input face hole Images Classification, and can comprise the condition relevant to external environment condition, for example, take posture, the angle of the attitude showing in input face hole pattern picture or Facial expression, shooting face image, be radiated at direction of light on face etc.For example, when according to Facial expression during by new input face hole Images Classification, this standard can be smile's face, amimia face, angry face, the face etc. of frowning, and based on these standards by face Images Classification.In addition, when when taking the angle of face by new input face hole Images Classification, whether this standard can take face from front, whether from the left side or the right take face, from front, face is taken on the slightly oblique left side or the right, whether from front, face is taken on the oblique left side or the right very largely, from front, slightly face is taken in top, from front, slightly face is taken in below, from front, significantly face is taken in top, from front, significantly face or no from positive upper left side is taken in below, lower left, face is taken in upper right side or lower right etc., and based on these standards by input face hole Images Classification.When according to image irradiation direction during by new input face hole Images Classification, this standard can be top or the lower edge direction etc. of the left side of back side direction, face of frontal, the face of face or the right direction, face, and with based on these standards by input face hole Images Classification.
Definite, in the situation of the input face hole pattern picture of people A, according to various features by input face hole Images Classification, and whether definite reference surface hole data storehouse 110 has stored the face image with the condition identical with the condition of classifying of input face hole pattern picture (for example, the attitude of input face hole pattern picture, Facial expression and lighting angle) of people A above.For example, in reference surface hole data storehouse 110, in the face image of the people A of registration, only comprise from the situation of the right and the positive face image of taking, if the face image of input is the image of taking from the face left side of people A recently, because (the data of face image of taking from the left side are not included in reference surface hole data storehouse 110, because taking the condition of the face of people A from the left side is the condition not comprising reference surface hole data storehouse 110), so can select this new input face hole pattern picture as the image that can additionally register in reference surface hole data storehouse 110, and therein by this image registration (S203).
In addition, not only the multiple aspect of these embodiment can be applied to above-described peculiar situation, can also be applied to following common situations.For example, when the class condition with reference to the face image of previous registration in reference surface hole data storehouse 110 is definite and the class condition of input face hole image correlation, and with the class condition of this input face hole image correlation be specific classification condition, can consider a kind of method: the face image relevant to this specific classification condition not as the predetermined threshold value in reference surface hole data storehouse looks like input face hole pattern to be chosen as when abundant can be in reference surface hole data storehouse the image of registration additionally.Herein, whether the face image relevant to specific classification condition be not as predetermined threshold value fully can only be determined and maybe can consider that ordinary people determines for specific people.
Fig. 3 uses according to the method for embodiment of the present disclosure by the schematic diagram that adds registration face selected cell 120 and select the process of the face image that will additionally register.
In Fig. 3, select the method for face image to illustrate that the angle based on taking face determines whether face image is the example of new face image.With reference to figure 3, when for people A, having face image a to the d(face image a in the previously stored face image in reference surface hole data storehouse 110 for facial recognition is positive picture, face image b is the picture to the face of the right rotation 45 degree, face image c is the picture to the face of left side rotation 45 degree, and face image d is the picture to the face of the right 90-degree rotation), if input face hole pattern similarly is the face image of people A and identical with face image e (face image e is the picture to the face of left side 90-degree rotation) recently, determining according to additional registration face selected cell 120 of the present disclosure that the shooting angle of the face image of the people A comprising in face image e belongs to from people A had previously been stored in the classification that the shooting angle of the face image in reference surface hole data storehouse 110 is different or had belonged to the identical category that comprises the shooting angle of the face image of storage in reference surface hole data storehouse 110.In this example, when determining that the shooting angle of new face image is included in that in the shooting angle classification of the face image of storage in reference surface hole data storehouse 110, (two shooting angle not necessarily match each other completely, if and they are only in identification when face within the scope of enough similarity), be defined as previously stored face image type in reference surface hole data storehouse 110.For example, if the hole pattern of input face recently of people A is not shown as f() be the picture to the face of the right rotation 50 degree, because this new face image in the scope similar to the face image b of previously stored people A in reference surface hole data storehouse 110, so additional registration face selected cell 120 can not registered recently input face hole pattern as f in reference surface hole data storehouse 110.
But, as shown in Figure 3, because input face hole pattern is that the face of people A is to the shooting angle of the photo of left side 90-degree rotation as the shooting angle of e recently, and be not therefore included in the shooting angle of face image of the people A previously having registered in reference surface hole data storehouse 110, so additional registration face selected cell 120 can add the new input face hole pattern of people A to reference surface hole data storehouse 110 as e.
Although in Fig. 3 only based on shooting angle by face Images Classification and mark new face image it is additionally registered, can be by one of them of shooting angle, Facial expression, illumination etc. being combined a bit to the filming apparatus for the face image selecting recently to register using according to embodiment of the present disclosure.This is because if direction of illumination difference, and as the filming apparatus of digital camera may be defined as different faces by the image of identical shooting angle, and correspondingly facial recognition probability may reduce.Therefore, if based on different condition determine wherein the multiple face image to each shooting angle or Facial expression application different light direction and corresponding to each condition by this multiple face recording image in reference surface hole data storehouse 110, facial recognition rate will be enhanced.
Fig. 4 is the process flow diagram that the method for the face image that will additionally register according to the selection of another embodiment of the present disclosure is shown.
In order to select to advise and to be registered as the face image of new face image, according to the additional registration face selected cell 120 of other embodiment of the present disclosure, a kind of method can be proposed: at least one candidate's face image (cluster) in groups based on specific facial recognition tolerance with reference to face database 110 outsides, then suggestion can be used in the representative face image of easily identifying these other face images automatically.
Although in the embodiment of Fig. 2, select the method for face image comprise by individual determine that new input face hole pattern looks like to determine whether will be by the face image that look like from the outside new input face hole pattern receiving of reference surface hole data 110 to be defined as to record reference surface hole data storehouse 110, but can also comprise from the multiple new face images of reference surface hole data storehouse 110 outside reception according to the selection face image of other embodiment of the present disclosure as shown in Figure 4, selection represents the face image of these new face images, and be the face image that will register in reference surface hole data storehouse 110 by selected face image suggestion.
With reference to figure 4, according to the first outside one or more new input face hole pattern pictures (S401) that receive from reference surface hole data storehouse 110 of the additional registration face selected cell 120 of other embodiment of the present disclosure.Then, additional registration face selected cell 120 is by looking like to be divided into many group similar images (S402) by the new input face hole pattern receiving based on facial recognition tolerance in groups by new input face hole pattern picture.Refer in groups the matching value of the facial recognition of calculating each image, and matching value based on calculating looks like these input face hole patterns to be grouped into the highly similar image of many groups or is grouped into the multiple series of images for example, between the descriptor of key point with short distance (, Euclidean distance).That is, the new input face hole pattern of transmission similarly is the image differing from one another, and different images need to be grouped into many groups face image of similar type, the i.e. illumination of the shooting angle of the attitude of similar type or Facial expression, similar type, similar type etc.
Then in the face image that, additional registration face selected cell 120 divides into groups, select the representative face image (S403) of the face image corresponding with barycenter image or average image as this group.This process means can the most fully represent the presentation graphics that looks like to be selected as this group by the index plane hole pattern of the group in the face image of similar image grouping.For example, can the representative face image that be chosen as this group close to the face image of the matching value of barycenter matching value or average will be there is in the matching value calculating in process in groups in each image.In other words, not whole identical images by the face image reality of positive face image packets, and there is different Facial expressions or slightly different shooting angle, and this can be to select the process that the standard picture of face image of grouping is presentation graphics.
Additional registration face selected cell 120 selects this type of face image as the image that can add the face image of previously having registered in reference surface hole data storehouse 110 to, and in reference surface hole data storehouse 110, registers these face images (S404).
Except above-described embodiment, can also use the example of following amendment.For example, in reference surface hole data storehouse 110 additionally when Registered Representative face image, with reference in reference surface hole data storehouse 110 previously the class condition of the face image of registration determine and the class condition of these representative face image correlations.If with the class condition of representative face image correlation be specific classification condition, only just can be by the additionally registration in reference surface hole data storehouse 110 of this representativeness face image when abundant not as the predetermined threshold value in reference surface hole data storehouse 110 at the face image relevant to this specific classification condition.
Register in advance the new input face hole pattern of selecting and delete quantity and interpolation in the face image that face replacement unit 130 can store from reference surface hole data storehouse 110 as many or quantity and guarantee in the required reference surface hole data storehouse 110 of the additional storage space expected that previously stored face image (, can say, look like to replace replacing with effective face data for facial recognition with the input face hole pattern of newly registering by the image of deleting previous registration).
Fig. 5 is the process flow diagram that the method for the face image (, the face image of deletion) that will be replaced by new input face hole pattern picture according to selection in the face image of previously having been registered from reference surface hole data storehouse 110 by the face replacement unit 130 of registering in advance of embodiment of the present disclosure is shown.
Can be chosen in the face image of storing maximum duration in reference surface hole data storehouse 110 as the image in the face image of the previous registration that will replace with new input face hole pattern picture according to the face replacement unit 130 of registration in advance of embodiment of the present disclosure.
With reference to figure 5, first grade with reference to previously stored face image in face database 110 by the order of storage according to the face replacement unit 130 of the registration in advance of embodiment of the present disclosure, to select the face image (S501) of the previous registration that will replace.In this example, can also suppose situation about quoting simply about the information of the fixed grading of order with storage.Then, can select the face image of (S502) the oldest previous registration, and by its deletion.In this example, can also replace with new input face hole pattern picture the face image (S503) of selected previous registration.
Fig. 6 is the process flow diagram illustrating according to the method for being selected the face image that will be replaced by new input face hole pattern picture by the face replacement unit 130 of registering in advance from the face image of previous registration of another embodiment of the present disclosure.
From the face image of previous registration, select the face image that will be replaced by new input face hole pattern picture can comprise by face image is counted to select and replaced to the matching times completing in successful facial recognition operation, and the face image of the previous registration with smallest match number of times is defined as recognition performance contribution minimum.
With reference to figure 6, in certain embodiments, first the face replacement unit 130 of registration is counted the number of times that in reference surface hole data storehouse 110, previously each face image of registration had been used to facial recognition in advance, to select the image (S601) of the previous registration that will replace.For example, if store 20 face images in reference surface hole data storehouse 110, put from Preset Time the number of times that current point in time is used to facial recognition with bulk billing system to each face image first to the 20 face image and count (target that, is used as similarity matching in reference surface hole data storehouse 110 is used to similarity matching to identify each face image how many times of the identity of input picture face recently).
Then, the face replacement unit 130 of registration can be selected to have for the face image (S602) of the previous registration of the minimum number of facial recognition and by its deletion in advance.At this some place, can also replace with new input face hole pattern picture the face image of (S603) selected previous registration.
That is, for example, although have the image of taking from top in reference surface hole data storehouse 110 in the face image of the previous registration of storage, the very rare image of taking from top in the image of recently taking.In other words, may there is such situation: the image of the previous registration of taking from top is very rarely used to facial recognition.In this type of situation, in advance the face replacement unit 130 of registration by this previously face image of registration to be defined as recognition performance contribution minimum, and by this previously face image of registration be chosen as the target that will delete or replace.
Fig. 7 is the process flow diagram illustrating according to the method for being selected the face image that will be replaced by new input face hole pattern picture by the face replacement unit 130 of registering in advance from the face image of previous registration of another embodiment of the present disclosure.
For the face image of selecting to replace with new input face hole pattern picture from the face image of previous registration, in certain embodiments, the face replacement unit 130 of registration can be by comparing face image and selecting one of the most similar face image to check the similarity between the face image of storage in reference surface hole data storehouse 110 as the face that will replace in advance.If there are two or more complete same face images, this may cause with only have one entirely with the situation of face image compared with, more spaces take reference surface hole data storehouse 110 in the identical facial recognition rate of maintenance in, and therefore determine one of them that can delete these complete same face images.
With reference to figure 7, first the method for the face image of the previous registration that selection will be replaced comprises the similarity degree (S701) of determining between the face image of previously having registered in reference surface hole data storehouse 110 and selects the most similar face image (S702).In this example, the process of determining similarity can use in the face image of registration in reference surface hole data storehouse 110 between the matching value of facial recognition poor (for example, demarcate the face image of people A) with the difference hour between matching value, determine that two similaritys between corresponding surface hole pattern picture are for high.Then, finally can be chosen as the image (S703) that will replace by being defined as one of two the most similar each other face images, and can be by this image-erasing.In this example, can also replace with new input face hole pattern picture the face image (S704) of selected previous registration.
Fig. 8 is the process flow diagram illustrating according to the method for being selected the face image that will be replaced by new input face hole pattern picture by the face replacement unit 130 of registering in advance from the face image of previous registration of another embodiment of the present disclosure.
For the face image of selecting to replace with new input face hole pattern picture from the face image of previous registration, according to other embodiment, in advance the face replacement unit 130 of registration can computing reference face database 110 in variation in the classification of face image of storage, and the face image with minimum change value is chosen as to the image that will replace.
Have again, in the method, if the same with the embodiment of Fig. 7, there are two or more complete same face images, this may cause with only have one entirely with the situation of face image compared with, more spaces take reference surface hole data storehouse 110 in the identical facial recognition rate of maintenance in, and therefore can determine one of them that can delete these complete same face images.But the face image by calculating variation in whole classification with minimum change value can be selected, and without one by one determining the similarity between the face image in reference surface hole data storehouse 110.
With reference to figure 8, according to the first previously variation (S801) in the classification between the face image of registration in computing reference face database 110 of the face replacement unit 130 of the registration in advance of other embodiment of the present disclosure, and select to have the face image (S802) of minimum change value to select the face image of the previous registration that will replace.At this some place, the process of calculating the variation in classification can be used known changing value computation process.Changing value can also obtain as follows: select one of face image of previously registration, calculate the mean value of the facial recognition matching value of the face image of the previous registration beyond selected face image, and calculate poor between the facial recognition matching value of face image of previous registration of a selection and the mean value of other facial recognition matching value.As alternative, can obtain as follows changing value: calculate difference between each facial recognition matching value and all facial recognition matching values are averaged instead of get rid of one of face image of previous registration.
Then, can replace selected image (S803) with new face image, and delete selected image.At this some place, can also replace with new input face hole pattern picture the face image of selected previous registration.
Selection new images is that the method for face image can be additionally registration as described in Fig. 2 and Fig. 4, selects the method for the image that will replace can combine and be used in many ways as Fig. 5 to Fig. 8 is described.
Above-describedly can adopt the form of program command to realize according to embodiment of the present disclosure, these program commands can be carried out and are recorded in computer readable recording medium storing program for performing by multiple computer module.This computer readable recording medium storing program for performing can comprise program command, data file, data structure etc. individually or in the mode of combination.The program command recording in computer readable recording medium storing program for performing can be specialized designs and be disposed for program command of the present disclosure or be known as the program command that the technician of computer software fields uses.This computer readable recording medium storing program for performing comprises, for example, magnetic medium, as hard disk, floppy disk and tape, optical recording media, as CD-ROM and DVD, magnet-optical medium, as CD, and special configuration becomes to store and the hardware unit of executive routine order, as ROM, RAM, flash memory etc.Program command comprises, for example, and the machine code that computing machine uses higher-level language code that interpreter etc. carries out and compiler to generate.Hardware unit can be configured to carry out work by one or more software modules, to carry out according to processing of the present disclosure, otherwise and.
In discussing above, although the disclosure is and certain content, if specific components, various embodiments and accompanying drawing are in conjunction with describing, they only provide in order to help to understand the disclosure, and the disclosure is not limited to these embodiment.By obvious, those skilled in the art can carry out multiple amendment and change to it according to these descriptions.
Therefore, spirit of the present disclosure should not be limited to above-described embodiment, and appended claims and it is equated or revises equivalently gained and will be considered as dropping in the scope of the present disclosure.
Claims (42)
1. for managing reference surface hole data storehouse for improve a method for facial recognition performance at limited memory environments, described method comprises:
Obtain new input face hole pattern picture;
With reference to the previously class condition of the face image of registration of the described reference surface hole data storehouse for facial recognition, determine the class condition with described input face hole image correlation; And
If to the described class condition of described input face hole image correlation be specific classification condition and with described specific classification condition relevant face image not as the predetermined threshold value in described reference surface hole data storehouse abundant, select described input face hole pattern picture as can be in described reference surface hole data storehouse the image of registration additionally.
2. the method for claim 1, in wherein said reference surface hole data storehouse previously the described class condition of the described face image of registration comprise described input face hole pattern picture Facial expression, take the angle of described face image, the direction of light projection on described face image and the type of light source at least one of them.
3. the method for claim 1, the wherein identity of the people based on comprising in described input face hole pattern picture, determines whether the described face image relevant to described specific classification condition is not so good as described predetermined threshold value in described reference surface hole data storehouse abundant.
4. the method for claim 1, be also included in described reference surface hole data storehouse, additionally register described input face hole pattern as time select and delete in described reference surface hole data storehouse wherein one or more of the previous described face image of registration.
5. method as claimed in claim 4, the step of wherein said deletion comprises:
In described reference surface hole data storehouse, previously in the described face image of registration, selected the oldest face image; And
Delete selected face image.
6. method as claimed in claim 4, the step of wherein said deletion comprises:
To in described reference surface hole data storehouse previously each face image of the described face image of registration put and be used to similarity matching till current and count to identify the number of times of the identity of the face comprising input picture recently from Preset Time;
In described reference surface hole data storehouse, previously in the described face image of registration, selected to there is the image for the minimum number of described similarity matching; And
Delete selected face image.
7. method as claimed in claim 4, the step of wherein said deletion comprises:
Determine the similarity between the described face image of previously having registered in described reference surface hole data storehouse;
Select previously between the face image of registration, there is the face image that approaches similarity most described in described reference surface hole data storehouse;
One of them of the selected face image of further selection; And
Delete the face image of further selecting.
8. method as claimed in claim 7, wherein in the step of described definite similarity, use in described reference surface hole data storehouse in the described face image of previously registration poor between the matching value of facial recognition, difference hour between described matching value, determines that two similaritys between corresponding surface hole pattern picture are for high.
9. method as claimed in claim 4, the step of wherein said deletion comprises:
Calculate the variation in the classification between the described face image of previously having registered in described reference surface hole data storehouse;
Selection has the face image of minimum change value; And
Delete selected face image.
10. method as claimed in claim 9, the step that wherein said calculating changes comprises:
Select in described reference surface hole data storehouse one of them of described face image of previously registration;
Calculate the mean value of the facial recognition matching value of the face image of the described previous registration except selected face image; And
Calculate poor between the facial recognition matching value of face image of a described selected previous registration and the mean value of described other facial recognition matching value.
11. methods as claimed in claim 9, the step that wherein said calculating changes comprises:
Calculate the mean value of the facial recognition matching value of the described face image of previously having registered in described reference surface hole data storehouse; And
Calculate poor between each facial recognition matching value of the described mean value of described facial recognition matching value and the face image of described previous registration.
12. 1 kinds for managing reference surface hole data storehouse for improve the method for facial recognition performance at limited memory environments, and described method comprises:
Obtain one or more new input face hole pattern pictures;
By described input face hole pattern picture being looked like described input face hole pattern to be grouped into many group similar images in groups based on specific face resolution amount;
In the input face hole pattern picture of described grouping, select the face image corresponding with barycenter face image or mean surface hole pattern picture as representative face image; And
Additionally register selected representative face image in the described reference surface hole data storehouse for facial recognition.
13. methods as claimed in claim 12, the step of wherein said grouping comprises:
Calculate the matching value for facial recognition in each input face hole pattern picture; And
Matching value based on calculated looks like to be grouped into many group height similar images by described input face hole pattern.
14. methods as claimed in claim 13, wherein at described selection face image in the step as representative face image, be chosen in the face image having in calculated matching value close to the matching value of barycenter matching value or mean match value as described representative face image.
15. methods as claimed in claim 12, the wherein said step of additionally registering selected representative face image in described reference surface hole data storehouse comprises:
With reference to the class condition of the described face image of previously having registered in described reference surface hole data storehouse, determine the class condition with described representative face image correlation; And
If with the described class condition of described representative face image correlation be specific classification condition, and the face image relevant to described specific classification condition not as the predetermined threshold value in described reference surface hole data storehouse abundant, by the additionally registration in described reference surface hole data storehouse of described representative face image.
16. methods as claimed in claim 12, be also included in described reference surface hole data storehouse, additionally register described input face hole pattern as time, select and delete in described reference surface hole data storehouse the previously some of them of the described face image of registration.
17. methods as claimed in claim 16, the step of wherein said deletion comprises:
From described reference surface hole data storehouse, previously in the described face image of registration, selected the oldest face image; And
Delete selected face image.
18. methods as claimed in claim 16, the step of wherein said deletion comprises:
To in described reference surface hole data storehouse previously each face image of the described face image of registration put till current and count to identify the number of times of the identity of the face that input picture comprises recently for similarity matching from Preset Time;
In described reference surface hole data storehouse, previously in the described face image of registration, selected to there is the image for the minimum number of described similarity matching; And
Delete selected face image.
19. methods as claimed in claim 16, the step of wherein said deletion comprises:
Determine the similarity between the described face image of previously having registered in described reference surface hole data storehouse;
Select to there is the face image that approaches similarity most between the face image of described previous registration;
Select one of them of selected face image; And
Delete selected face image.
20. methods as claimed in claim 19, the step of wherein said definite similarity comprises, use in described reference surface hole data storehouse in the described face image of previously registration poor between the described matching value of facial recognition, difference between described matching value more hour, determines that two similaritys between corresponding surface hole pattern picture are higher.
21. methods as claimed in claim 16, the step of wherein said deletion comprises:
Calculate the variation in the classification between the described face image of previously having registered in described reference surface hole data storehouse;
Selection has the face image of minimum change value; And
Delete selected face image.
22. methods as claimed in claim 21, the step that wherein said calculating changes comprises:
Select in described reference surface hole data storehouse one of them of described face image of previously registration;
Calculate the mean value of the facial recognition matching value of the face image of the described previous registration except selected face image; And
Calculate poor between the facial recognition matching value of face image of a described selected previous registration and the described mean value of described other facial recognition matching value.
23. methods as claimed in claim 21, the step that wherein said calculating changes comprises:
Calculate the mean value of the facial recognition matching value of the described face image of previously having registered in described reference surface hole data storehouse; And
Calculate poor between each facial recognition matching value of the described mean value of described facial recognition matching value and the face image of described previous registration.
24. 1 kinds for improving the reference surface hole data library management device of facial recognition performance at limited memory environments, described device comprises:
Reference surface hole data storehouse, described reference surface hole data storehouse is configured to the face image of record for facial recognition; And
Additional registration face selected cell, described additional registration face selected cell is configured to, if need input face hole pattern picture, the class condition of the face image that reference had previously been registered in described reference surface hole data storehouse, determine the class condition with described new input face hole image correlation, and be configured to, if to the described class condition of described input face hole image correlation be specific classification condition and with described specific classification condition relevant face image not as the predetermined threshold value in described reference surface hole data storehouse abundant, select described input face hole pattern picture as the image that can additionally register in described reference surface hole data storehouse.
25. devices as claimed in claim 24, Facial expression that wherein said reference surface hole data storehouse comprises described input face hole pattern picture, take the angle of described face image, the direction of light projection on described face image and the type of light source at least one of them, as the described class condition of the face image of described previous registration.
26. devices as claimed in claim 24, wherein said additional registration face selected cell is configured to the identity of the people based on comprising in described input face hole pattern picture, determines whether the described face image relevant to described specific classification condition is not so good as described predetermined threshold value in described reference surface hole data storehouse abundant.
27. devices as claimed in claim 24, also comprise the face replacement unit of registration in advance, the described face replacement unit of registration be in advance configured to additionally register in described reference surface hole data storehouse described input face hole pattern as time, select and delete in described reference surface hole data storehouse the previously some of them of the described face image of registration.
28. devices as claimed in claim 27, the wherein said face replacement unit of registration is in advance configured to previously in the described face image of registration, select and deleted the oldest face image from described reference surface hole data storehouse.
29. devices as claimed in claim 27, the wherein said face replacement unit of registration be in advance configured to in described reference surface hole data storehouse previously each face image of the described face image of registration from Preset Time point till currently count to identify the number of times of the identity of the face that input picture comprises recently for similarity matching; And be configured to previously in the described face image of registration, select and deleted the image having for the minimum number of described similarity matching in described reference surface hole data storehouse.
30. devices as claimed in claim 27, the wherein said face replacement unit of registration is in advance configured to determine in described reference surface hole data storehouse the previously similarity between the described face image of registration, and selects and delete between the face image of described previous registration to have the face image that approaches similarity most.
31. devices as claimed in claim 30, wherein in the time determining described similarity, the described face replacement unit of registration is in advance configured to, use in described reference surface hole data storehouse in the described face image of previously registration poor between the described matching value of facial recognition, difference hour between described matching value, determines that two similaritys between corresponding surface hole pattern picture are for high.
32. devices as claimed in claim 27, the wherein said face replacement unit of registration is in advance configured to calculate the variation in the classification between the described face image of previously having registered in described reference surface hole data storehouse, and selects and delete the face image with minimum change value.
33. devices as claimed in claim 32, wherein in the time calculating described changing value, the described face replacement unit of registration is in advance configured to select one of face image of previously having registered in described reference surface hole data storehouse, calculate the mean value of the facial recognition matching value of the face image of the described previous registration beyond selected face image, and calculate poor between the facial recognition matching value of face image of previous registration of a selection and the mean value of described other facial recognition matching value.
34. devices as claimed in claim 32, wherein in the time calculating described variation, the described face replacement unit of registration is in advance configured to calculate in described reference surface hole data storehouse the previously mean value of the facial recognition matching value of the described face image of registration, and calculates poor between each facial recognition matching value of the described mean value of described facial recognition matching value and the face image of described previous registration.
35. 1 kinds for improving the reference surface hole data library management device of facial recognition performance at limited memory environments, described device comprises:
Reference surface hole data storehouse, described reference surface hole data storehouse is configured to the face image of record for facial recognition; And
Additional registration face unit, described additional registration face cell location becomes, if need input face hole pattern picture, by described input face hole pattern picture being looked like one or more new input face hole patterns to be grouped into many group similar images in groups based on specific face resolution amount, and be configured to select the face image corresponding with barycenter face image or mean surface hole pattern picture as representative face image in the input face hole pattern picture of described grouping, and by the additionally registration in described reference surface hole data storehouse of selected representative face image.
36. devices as claimed in claim 35, wherein said additional registration face selected cell is configured to calculate the matching value for facial recognition in each input face hole pattern picture, and matching value based on calculated looks like to be divided into many group height similar images by described input face hole pattern.
37. devices as claimed in claim 36, wherein said additional registration face selected cell is configured to, and is chosen in the face image having in calculated matching value close to the matching value of barycenter matching value or mean match value as described representative face image.
38. devices as claimed in claim 35, while wherein additionally registering selected representative face image in described reference surface hole data storehouse, the class condition of the face image of previously having registered in described additional registration face selected cell reference described reference surface hole data storehouse is determined the class condition with described representative face image correlation, if and with the class condition of described representative face image correlation be specific classification condition, and the face image relevant to described specific classification condition not as the predetermined threshold value in described reference surface hole data storehouse abundant, by the additionally registration in described reference surface hole data storehouse of described representative face image.
39. devices as claimed in claim 35, also comprise the face replacement unit of registration in advance, the face replacement unit of described registration in advance for additionally register in described reference surface hole data storehouse described input face hole pattern as time, select and delete in described reference surface hole data storehouse wherein one or more of described face image of previously registration.
40. devices as claimed in claim 39, the wherein said face replacement unit of registration be in advance configured to in described reference surface hole data storehouse previously each face image of the described face image of registration from Preset Time point till current for similarity matching so that identifying the number of times of the identity of the face that input picture comprises recently counts, and be configured to previously in the described face image of registration, select and deleted the image having for the minimum number of described similarity matching in described reference surface hole data storehouse.
41. devices as claimed in claim 39, the wherein said face replacement unit of registration is in advance configured to calculate the variation in the classification between the described face image of previously having registered in described reference surface hole data storehouse, and selects and delete the face image with minimum change value.
42. 1 kinds of computer readable recording medium storing program for performing for logger computer program, described computer program is for selecting and replace face image according to the method described in claim 1 to 23 any one.
Applications Claiming Priority (3)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| KR10-2011-0097796 | 2011-09-27 | ||
| KR1020110097796A KR101180471B1 (en) | 2011-09-27 | 2011-09-27 | Method, apparatus and computer-readable recording medium for managing reference face database to improve efficiency of face recognition using small-capacity memory |
| PCT/KR2012/007854 WO2013048162A2 (en) | 2011-09-27 | 2012-09-27 | Method, apparatus and computer readable recording medium for managing a reference face database to improve face recognition performance under a restricted memory environment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104081415A true CN104081415A (en) | 2014-10-01 |
| CN104081415B CN104081415B (en) | 2017-10-20 |
Family
ID=47073942
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201280058022.2A Active CN104081415B (en) | 2011-09-27 | 2012-09-27 | For managing plane of reference hole data storehouse for method, device and the computer readable recording medium storing program for performing of the raising facial recognition performance under limited memory environments |
Country Status (6)
| Country | Link |
|---|---|
| US (1) | US9298976B2 (en) |
| EP (1) | EP2767930A4 (en) |
| JP (1) | JP5936697B2 (en) |
| KR (1) | KR101180471B1 (en) |
| CN (1) | CN104081415B (en) |
| WO (1) | WO2013048162A2 (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110023992A (en) * | 2016-12-05 | 2019-07-16 | 索尼互动娱乐股份有限公司 | System, fixture, information processing unit, information processing method and program |
| CN110023993A (en) * | 2016-12-05 | 2019-07-16 | 索尼互动娱乐股份有限公司 | Information processing equipment, information processing method and program |
| CN114758530A (en) * | 2022-04-28 | 2022-07-15 | 浙江理工大学 | Infant face capability training program and training method |
Families Citing this family (38)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5617627B2 (en) * | 2010-12-28 | 2014-11-05 | オムロン株式会社 | Monitoring device and method, and program |
| JP5919665B2 (en) * | 2011-07-19 | 2016-05-18 | 日本電気株式会社 | Information processing apparatus, object tracking method, and information processing program |
| KR101130817B1 (en) * | 2011-09-27 | 2012-04-16 | (주)올라웍스 | Face recognition method, apparatus, and computer-readable recording medium for executing the method |
| KR101297736B1 (en) * | 2013-05-07 | 2013-08-23 | 주식회사 파이브지티 | Face recognition method and system |
| KR101316805B1 (en) * | 2013-05-22 | 2013-10-11 | 주식회사 파이브지티 | Automatic face tracing and recognition method and system |
| JP5902661B2 (en) * | 2013-09-30 | 2016-04-13 | 株式会社東芝 | Authentication apparatus, authentication system, and authentication method |
| JP6268960B2 (en) * | 2013-11-15 | 2018-01-31 | オムロン株式会社 | Image recognition apparatus and data registration method for image recognition apparatus |
| JP6427973B2 (en) | 2014-06-12 | 2018-11-28 | オムロン株式会社 | Image recognition apparatus and feature data registration method in image recognition apparatus |
| US9864900B2 (en) * | 2014-06-26 | 2018-01-09 | Cisco Technology, Inc. | Entropy-reducing low pass filter for face-detection |
| US9384385B2 (en) * | 2014-11-06 | 2016-07-05 | Intel Corporation | Face recognition using gradient based feature analysis |
| CN105989352B (en) | 2015-03-06 | 2019-08-20 | 华为技术有限公司 | Image recognition accelerator, terminal device and image-recognizing method |
| US9619696B2 (en) * | 2015-04-15 | 2017-04-11 | Cisco Technology, Inc. | Duplicate reduction for face detection |
| CN105138962A (en) * | 2015-07-28 | 2015-12-09 | 小米科技有限责任公司 | Image display method and image display device |
| US10769255B2 (en) | 2015-11-11 | 2020-09-08 | Samsung Electronics Co., Ltd. | Methods and apparatuses for adaptively updating enrollment database for user authentication |
| CN106706677B (en) * | 2015-11-18 | 2019-09-03 | 同方威视技术股份有限公司 | Check the method and system of cargo |
| CN105938552B (en) * | 2016-06-29 | 2020-04-24 | 北京旷视科技有限公司 | Face recognition method and device for automatically updating base map |
| US10579940B2 (en) | 2016-08-18 | 2020-03-03 | International Business Machines Corporation | Joint embedding of corpus pairs for domain mapping |
| US10642919B2 (en) | 2016-08-18 | 2020-05-05 | International Business Machines Corporation | Joint embedding of corpus pairs for domain mapping |
| US10657189B2 (en) | 2016-08-18 | 2020-05-19 | International Business Machines Corporation | Joint embedding of corpus pairs for domain mapping |
| US10621419B2 (en) * | 2016-12-06 | 2020-04-14 | Robert William Kocher | Method and system for increasing biometric acceptance rates and reducing false accept rates and false rates |
| KR102233175B1 (en) | 2017-01-05 | 2021-03-29 | 한국전자통신연구원 | Method for determining signature actor and for identifying image based on probability of appearance of signature actor and apparatus for the same |
| WO2018148037A1 (en) * | 2017-02-10 | 2018-08-16 | Selfiepay, Inc. | Systems and methods for biometric transaction management |
| JP7299543B2 (en) * | 2017-06-30 | 2023-06-28 | キヤノンマーケティングジャパン株式会社 | Information processing system, information processing method, program |
| CN108228871A (en) * | 2017-07-21 | 2018-06-29 | 北京市商汤科技开发有限公司 | Facial image dynamic storage method and device, electronic equipment, medium, program |
| CN108875476B (en) * | 2017-08-03 | 2022-03-15 | 北京旷视科技有限公司 | Automatic near-infrared face registration and recognition method, device and system and storage medium |
| US10489690B2 (en) * | 2017-10-24 | 2019-11-26 | International Business Machines Corporation | Emotion classification based on expression variations associated with same or similar emotions |
| KR102466942B1 (en) | 2017-12-27 | 2022-11-14 | 한국전자통신연구원 | Apparatus and method for registering face posture for face recognition |
| CN109977746B (en) * | 2017-12-27 | 2023-11-24 | 韩国电子通信研究院 | Apparatus and method for registering facial gestures for facial recognition |
| US10819789B2 (en) | 2018-06-15 | 2020-10-27 | At&T Intellectual Property I, L.P. | Method for identifying and serving similar web content |
| CN111091529A (en) * | 2018-10-24 | 2020-05-01 | 株式会社理光 | People counting method and system |
| US20200151453A1 (en) * | 2018-11-08 | 2020-05-14 | International Business Machines Corporation | Reducing overlap among a collection of photographs |
| CN111325059A (en) * | 2018-12-14 | 2020-06-23 | 技嘉科技股份有限公司 | Face recognition method, device and computer readable medium |
| TWI682329B (en) | 2018-12-14 | 2020-01-11 | 技嘉科技股份有限公司 | Method, device and computer readable medium of facial recognition |
| US10635918B1 (en) * | 2019-01-30 | 2020-04-28 | StradVision, Inc. | Method and device for managing smart database for face recognition based on continual learning |
| US10621473B1 (en) * | 2019-01-30 | 2020-04-14 | StradVision, Inc. | Method for providing object detecting system capable of updating types of detectable classes in real-time by using continual learning and devices using the same |
| DE112019006940T5 (en) * | 2019-02-28 | 2021-11-25 | Google Llc | User authentication through pose-based facial recognition |
| CN109948494B (en) * | 2019-03-11 | 2020-12-29 | 深圳市商汤科技有限公司 | Image processing method and device, electronic equipment and storage medium |
| WO2023166553A1 (en) * | 2022-03-01 | 2023-09-07 | 日本電気株式会社 | Update notification device, authentication device, update notification method, and computer-readable medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1374617A (en) * | 2001-03-05 | 2002-10-16 | 欧姆龙株式会社 | Oranism checking apparatus, system and method, and renew method of registered data |
| US20060204058A1 (en) * | 2004-12-07 | 2006-09-14 | Kim Do-Hyung | User recognition system and method thereof |
| US20090028393A1 (en) * | 2007-07-24 | 2009-01-29 | Samsung Electronics Co., Ltd. | System and method of saving digital content classified by person-based clustering |
| US20090141949A1 (en) * | 2007-12-03 | 2009-06-04 | Samsung Electronics Co., Ltd. | Method and apparatus for recognizing a plural number of faces, and method and apparatus for registering face, and an image capturing method and system |
| US20090185723A1 (en) * | 2008-01-21 | 2009-07-23 | Andrew Frederick Kurtz | Enabling persistent recognition of individuals in images |
Family Cites Families (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH07262379A (en) * | 1994-03-18 | 1995-10-13 | N T T Data Tsushin Kk | Article identifying system |
| JPH10275236A (en) * | 1997-03-28 | 1998-10-13 | Oki Electric Ind Co Ltd | Animal identification system |
| JP4153691B2 (en) * | 2001-11-06 | 2008-09-24 | 株式会社東芝 | Face image matching device and face image matching method |
| JP2004342073A (en) * | 2003-04-25 | 2004-12-02 | Nec Corp | Fingerprint authentication system and method |
| KR100571826B1 (en) * | 2003-12-02 | 2006-04-17 | 삼성전자주식회사 | Large volume face recognition appratus and method thereof |
| US7627406B2 (en) * | 2005-01-13 | 2009-12-01 | General Motors Corporation | System and method for data storage and diagnostics in a portable communications device interfaced with a telematics unit |
| JP2006313428A (en) * | 2005-05-06 | 2006-11-16 | Toyota Motor Corp | Database evaluation device |
| JP5010905B2 (en) * | 2006-12-13 | 2012-08-29 | パナソニック株式会社 | Face recognition device |
| JP4336729B2 (en) * | 2007-03-28 | 2009-09-30 | シャープ株式会社 | Image processing apparatus, image forming apparatus, image transmitting apparatus, image reading apparatus, image processing system, image processing program, and recording medium thereof |
| KR101363017B1 (en) * | 2007-08-23 | 2014-02-12 | 삼성전자주식회사 | System and methed for taking pictures and classifying the pictures taken |
| KR100898766B1 (en) * | 2007-09-21 | 2009-05-20 | 한국전자통신연구원 | Face recognition system, the method of face recognition registration and confirmation |
| JP5123037B2 (en) * | 2008-04-16 | 2013-01-16 | パナソニック株式会社 | Face authentication apparatus and face authentication method |
| JP2010027035A (en) | 2008-06-16 | 2010-02-04 | Canon Inc | Personal authentication equipment and personal authentication method |
| KR101185243B1 (en) * | 2009-12-15 | 2012-09-21 | 삼성전자주식회사 | Apparatus and method for registering a plurlity of face image for face recognition |
-
2011
- 2011-09-27 KR KR1020110097796A patent/KR101180471B1/en not_active IP Right Cessation
-
2012
- 2012-09-27 JP JP2014531738A patent/JP5936697B2/en active Active
- 2012-09-27 US US14/129,628 patent/US9298976B2/en active Active
- 2012-09-27 WO PCT/KR2012/007854 patent/WO2013048162A2/en active Application Filing
- 2012-09-27 EP EP12835834.8A patent/EP2767930A4/en not_active Withdrawn
- 2012-09-27 CN CN201280058022.2A patent/CN104081415B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1374617A (en) * | 2001-03-05 | 2002-10-16 | 欧姆龙株式会社 | Oranism checking apparatus, system and method, and renew method of registered data |
| US20060204058A1 (en) * | 2004-12-07 | 2006-09-14 | Kim Do-Hyung | User recognition system and method thereof |
| US20090028393A1 (en) * | 2007-07-24 | 2009-01-29 | Samsung Electronics Co., Ltd. | System and method of saving digital content classified by person-based clustering |
| US20090141949A1 (en) * | 2007-12-03 | 2009-06-04 | Samsung Electronics Co., Ltd. | Method and apparatus for recognizing a plural number of faces, and method and apparatus for registering face, and an image capturing method and system |
| US20090185723A1 (en) * | 2008-01-21 | 2009-07-23 | Andrew Frederick Kurtz | Enabling persistent recognition of individuals in images |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110023992A (en) * | 2016-12-05 | 2019-07-16 | 索尼互动娱乐股份有限公司 | System, fixture, information processing unit, information processing method and program |
| CN110023993A (en) * | 2016-12-05 | 2019-07-16 | 索尼互动娱乐股份有限公司 | Information processing equipment, information processing method and program |
| CN114758530A (en) * | 2022-04-28 | 2022-07-15 | 浙江理工大学 | Infant face capability training program and training method |
| CN114758530B (en) * | 2022-04-28 | 2023-08-08 | 浙江理工大学 | Infant face ability training program and training method |
Also Published As
| Publication number | Publication date |
|---|---|
| KR101180471B1 (en) | 2012-09-07 |
| WO2013048162A2 (en) | 2013-04-04 |
| EP2767930A2 (en) | 2014-08-20 |
| EP2767930A4 (en) | 2016-03-16 |
| US20140226877A1 (en) | 2014-08-14 |
| JP2014526755A (en) | 2014-10-06 |
| WO2013048162A3 (en) | 2013-05-23 |
| US9298976B2 (en) | 2016-03-29 |
| CN104081415B (en) | 2017-10-20 |
| JP5936697B2 (en) | 2016-06-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104081415A (en) | Method, apparatus and computer readable recording medium for managing a reference face database to improve face recognition performance under a restricted memory environment | |
| Liu et al. | Adaptive normalized representation learning for generalizable face anti-spoofing | |
| CN103189922B (en) | The method that data processing equipment performs and gather the system of metadata of duplicate pictures | |
| YuanQiang et al. | Guided attention network for object detection and counting on drones | |
| CN104426841A (en) | Method for arranging background image, and correlation server and system | |
| US20170323149A1 (en) | Rotation invariant object detection | |
| CN114329028A (en) | Data processing method, data processing equipment and computer readable storage medium | |
| CN116958267B (en) | Pose processing method and device, electronic equipment and storage medium | |
| CN114037912A (en) | Method and device for detecting change of remote sensing image and computer readable storage medium | |
| CN102792675A (en) | Method, system, and computer-readable recording medium for adaptively performing image-matching according to conditions | |
| Wang et al. | Cross-domain learning using optimized pseudo labels: toward adaptive car detection in different weather conditions and urban cities | |
| CN111178455B (en) | Image clustering method, system, device and medium | |
| CN111368128A (en) | Target picture identification method and device and computer readable storage medium | |
| US20220414393A1 (en) | Methods and Systems for Generating Composite Image Descriptors | |
| CN115880754A (en) | Multi-gear combination method and device and electronic equipment | |
| US20230196831A1 (en) | Image Group Classifier in a User Device | |
| CN112257666B (en) | Target image content aggregation method, device, equipment and readable storage medium | |
| CN114898182A (en) | Image data screening method and system based on target detection learning algorithm | |
| US11176696B2 (en) | Point depth estimation from a set of 3D-registered images | |
| Lin et al. | Realtime Vehicle Tracking Method Based on YOLOv5+ DeepSORT | |
| Wu et al. | Application of Intelligent Analysis Technology of Football Video Based on Online Target Tracking Algorithm of Motion Characteristics in Football Training | |
| US20230343095A1 (en) | Group Classifier Training Using Video Object Tracker | |
| US20230343094A1 (en) | Video Group Classification Using Object Tracker | |
| US11893012B1 (en) | Content extraction using related entity group metadata from reference objects | |
| US20230196832A1 (en) | Ranking Images in an Image Group |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |